Skip to content

01-understanding-load-average.md

Latest commit

 

History

History
64 lines (44 loc) · 3.98 KB

01-understanding-load-average.md

File metadata and controls

64 lines (44 loc) · 3.98 KB

Understanding Load Average in Android

In the world of Android optimization, "Load Average" is one of the most misunderstood metrics. Many users—and even some developers—equate high load strictly with high CPU usage. While correlated, they are not the same.

This document dives deep into the Linux kernel (which powers Android) to explain what "Load" actually represents and why it is the single most critical metric for system stability, battery life, and performance.

1. What is "Load"?

Technically, the system load average is a measurement of the computational work the system is performing. In the Linux kernel, the load average represents the number of processes that are either:

  1. Running: Currently executing on the CPU.
  2. Runnable: Waiting in the run queue for a CPU core to become available.
  3. Uninterruptible Sleep (D State): Waiting for a hardware resource (typically I/O) to respond.

Note

Unlike other Unix systems, Linux includes processes in Uninterruptible Sleep (disk I/O) in the load calculation. This is why a device can have a high load average even with 0% CPU usage.

2. The CPU Run Queue

Think of the CPU as a checkout counter at a supermarket.

  • Running process: The person currently paying.
  • Runnable process: People waiting in line.

If you have a Quad-Core CPU (4 checkout counters) and a load average of 4.0, it means all cores are busy, but no one is waiting. The system is at 100% saturation but running smoothly. If the load is 8.0 on a Quad-Core CPU, it means 4 processes are running, and 4 are waiting. This causes latency (lag).

The Android Context

Android devices often have 8 cores (e.g., 4 little, 3 big, 1 prime).

  • Load < 4.0: Ideally, background tasks should stay on the efficient "little" cores.
  • Load > 8.0: The system is saturated. Tasks are fighting for CPU time, causing UI jank (dropped frames).

3. The Silent Killer: Uninterruptible Sleep (D-State)

This is the most critical concept for Android diagnostics.

A process enters "Uninterruptible Sleep" (shown as D in top or ps) when it requests data from hardware and cannot proceed until it gets an answer. It cannot be killed or interrupted during this time.

Common Causes of D-State on Android:

  • Slow Storage (I/O Wait): Reading/Writing to the internal UFS/eMMC storage. If the storage controller is overwhelmed (e.g., by a backup process or a rogue app logging excessively), processes stack up in the queue waiting for disk access.
  • Kernel Locks: Drivers waiting for hardware interrupts.

Scenario: Your CPU usage is low (10%), but your phone feels incredibly sluggish, and the load average is 15.0.

  • Diagnosis: This is almost certainly I/O Wait. The CPU is idle because it's waiting for the storage to finish writing data.

4. Why High Load destroys Battery Life

High load triggers a vicious cycle known as the Thermal Death Spiral:

  1. High Load detected: The kernel's governor sees a long run queue.
  2. Frequency Spike: To clear the queue, the governor boosts CPU frequencies to maximum.
  3. Heat Generation: High frequencies generate massive heat.
  4. Thermal Throttling: The hardware hits thermal limits and drastically reduces CPU frequency to cool down.
  5. Load Increases: With lower frequency, processes take longer to finish, causing the run queue (Load) to grow even larger.
  6. Loop: The system is now stuck in a high-load, low-performance, high-heat state.

Summary

Metric Represents Impact
CPU Usage % How busy the processor is now. High usage drains battery, but doesn't always mean lag.
Load Average How many tasks are demanding resources (CPU + I/O). High load always means contention and potential lag.
I/O Wait Time spent waiting for disk/network. The invisible cause of UI freezes and "phantom" load.

Next Step: Learn how to measure these metrics using CLI Diagnostic Tools.